Forest fire susceptibility zonation of Mizoram using GIS: A comparative analysis of bivariate models

Abstract

Forest fires pose a significant environmental threat in Mizoram, a hilly region in northeast India, contributing to forest degradation, biodiversity loss, and alterations in the regional climate. The present study aimed to prepare forest fire susceptibility (FFS) zones of the state using Frequency Ratio (FR), Evidential Belief Function (EBF), and Index of Entropy (IOE), and their comparative analysis within a Geographic Information System (GIS) environment. The 453 fire incidents were classified into training (248, 70 %), used for model building and testing (105, 30 %) for validation. A total of eleven conditioning factors from different data sources were prepared for evaluation. The FFS was classified into categories from very low to very high using the quantile method. Model validation was performed using ROC–AUC (Receiver Operating Curve-Area Under Curve) metrics, revealing that the EBF model outperformed than FR and IOE with AUC scores of 0.92 (training) and 0.89 (testing). The results identified climatic factors and vegetation, particularly NDVI, as the most significant contributors to FFS. The novelty of this study lies in the combined application of FR, EBF, and IOE models, enabling a comprehensive comparison of their predictive performance and spatial accuracy. Furthermore, improved factor weighting and validation methods provide more reliable spatial insights for forest fire risk management. The results indicate that the majority of high-risk areas are concentrated in steep, sun-exposed southwestern slopes and areas near roads and settlements, where fuel conditions and human activities heighten ignition potential, demonstrating the effectiveness of the integrated modelling approach. These findings provide valuable insights for early fire detection, ecological modelling, land use management, and sustainable forest resource practices. By demonstrating a replicable, data-driven modelling approach in a data-scarce region, this study advances existing FFS research and lays the groundwork for dynamic forecasting and targeted risk mitigation strategies.